Society and Natural Resources, 23:58–70 Copyright # 2010 Taylor & Francis Group, LLC ISSN: 0894-1920 print=1521-0723 online DOI: 10.1080/08941920802179766
Insights and Applications Homebuyers and Wildfire Risk: A Colorado Springs Case Study PATRICIA ANN CHAMP USDA Forest Service, Rocky Mountain Research Station, Fort Collins, Colorado, USA
GEOFFREY H. DONOVAN USDA Forest Service, Pacific Northwest Research Station, Portland, Oregon, USA
CHRISTOPHER M. BARTH Montrose Interagency Fire Management Unit, Montrose, Colorado, USA In recent years, the threat that wildfire poses to homes has received much attention in both the mainstream press and academic literature. However, little is known about how homebuyers consider wildfire risk during the home-purchase process. In the context of a unique wildfire education program, we consider two approaches to examining the relationship between wildfire risk and home purchases. Results from a market-level analysis using home sales price data are compared to household survey results. The household survey validates the market-level analysis and provides further insight into homebuyers and wildfire risk. Specifically, we find that while homebuyers prefer locations near dangerous topography, they also prefer less flammable building materials. However, most homebuyers were unaware of wildfire risk when they made their home-purchase decisions. Keywords economics, environmental attitudes and concerns, hedonic price model, homebuyers, human behavior in the environment, survey methods, wildfire
Recent severe wildfire seasons in the western United States have increased public awareness of the dangers of wildfire. In particular, concern has focused on the wildland–urban interface, where homes abut forested lands and fuel loads are often elevated from decades of aggressive wildfire suppression. These increased fuel loads, along with years of drought and record-setting heat, have significantly increased the risk Received 31 October 2007; accepted 29 January 2008. Address correspondence to Patricia Ann Champ, USDA Forest Service, Rocky Mountain Research Station, 2150-A Centre Avenue, Fort Collins, CO 80526, USA. E-mail:
[email protected]
58
Homebuyers and Wildfire Risk
59
of wildfires. In addition, migration to areas proximate to national forests in the western United States has increased the numbers of homes at risk from wildfire (New York Times, June 26, 2007; Hammer et al. 2007). One policy response has been to institute programs that make homes less susceptible to wildfire by encouraging, or requiring, homeowners to use more fire-resistant building materials and to remove flammable vegetation from around their homes. Such programs have a common focus: They target people currently living in fire-prone areas. However, given the influx of people into fire-prone areas, it would seem appropriate to target potential as well as existing residents of fire-prone areas. Unfortunately, there is little information on the role of wildfire risk on home purchase decisions in fire-prone areas. Such information is crucial if policymakers wish to target potential, not just current, residents of fire-prone areas. We present the results of a household survey of recent homebuyers in the fire-prone area of Colorado Springs, CO. We chose Colorado Springs for two reasons. First, in 2002, the City of Colorado Springs Fire Department instituted a comprehensive wildfire education campaign. As part of this unique campaign, parcel-level wildfire-risk ratings were estimated for every address in the fire-prone area of the city. Second, a previous hedonic price study (Donovan, Champ, and Butry 2007) examined the effect of wildfire risk on the housing market in Colorado Springs before and after the wildfire risk ratings were made available on the Web. The study results suggested that homebuyers’ preferences with respect to wildfire risk were affected by the Colorado Springs Fire Department’s FireWise program but left many questions unanswered. The household survey was designed to both validate and complement the Donovan, Champ, and Butry study by providing a richer understanding of the role of wildfire risk in home purchase decisions. Specifically, the survey provides insight into how informed homebuyers are about wildfire risk, the level of concern about wildfire risk during the home purchase process, and how this concern is related to past experience with wildfire. In the next section we summarize the relevant literature on wildfire risk and homebuyers. Then we describe the study area, the Colorado Springs Fire Department’s FireWise program, and the data. In the fourth section, we summarize the results of the hedonic price analysis. We then describe the household survey data. In the last section, we discuss the study results.
Wildfire and Homebuyers The literature on wildfire risk and home purchases is thin. To our knowledge the Donovan, Champ, and Butry (2007) study is the only study to have directly estimated the impact of wildfire risk on housing prices. Loomis (2004) examined the effect of a large wildfire event on housing prices in a community that was 2 miles from the fire. By looking at housing prices 3 years before the wildfire and 5 years after the wildfire, Loomis found a significant drop in post-fire housing prices in the community proximate to the wildfire. Similar to the Loomis study, Mueller, Loomis, and Gonza´lez-Caba´n (2009) considered the effect of wildfire events on home prices. They found that multiple forest fires cause home prices to decrease in the area near the fires. Unlike other natural disasters such as a flood or a hurricane, a wildfire event would likely reduce wildfire risk and we might expect an increase in housing sales prices if homebuyers are risk adverse and risk was their dominant concern. However, the observed decrease in home purchase prices after a wildfire event suggests homebuyers are probably responding to the visual impacts of wildfires.
60
P. A. Champ et al.
Two studies, one on wildfire risk and one on flood risk, have focused on risk and home purchase decisions by directly surveying home owners. Vogt (2004) surveyed wildland–urban interface homeowners in California, Colorado, and Florida. Survey participants were asked to rate the level of consideration given to 11 structure and property attributes in the home purchase decision. Fire protection services such as firefighters and fire trucks received the highest average rating among all the attributes considered, while nonflammable roofing materials had the fourth highest rating for California and Colorado. The lot having few highly flammable trees had a relatively low rating in all three states. Chivers and Flores (2002) surveyed homeowners about their home purchase decision in a flood-prone area of Boulder, CO, and found that the individuals living in the special flood hazard area reported not understanding the risk when they purchased their homes. They found that only 8% of their study respondents learned of the potential flood risk associated with their home prior to making an offer on the home.
An Education Program in Colorado Springs Colorado Springs is a city of 361,000 on the front range of the Rocky Mountains in Colorado, approximately 70 miles south of Denver. The study area covers 45 square miles on the western edge of the city bordered by the Pike National Forest, the Air Force Academy, and the Fort Carson Army Base. The elevation in this area varies between 6,000 and 6,800 feet, and the mean annual precipitation is 15 inches. The intermixed forest is predominantly ponderosa pine (Pinus ponderosa) and gambel oak (Quercus gambelii) with some Douglas-fir (Pseudotsuga menziesii var. glauca), particularly at higher elevations. The area has a mixed-severity fire regime: Fires can vary from ground fires that cause little or no overstory mortality to stand-replacing fires. In an average year, the 240,000-acre Pikes Peak Ranger District of the Pike National Forest, which borders the study area, experiences between 40 and 50 wildfire ignitions. However, very few of these ignitions exceed five acres because they are either suppressed by fire crews or because the rain that typically accompanies lightning in this area puts them out naturally. Since European settlement, the study area has experienced two major fires. In 1854, a fire started approximately 7 miles southwest of downtown Colorado Springs on Cheyenne Mountain and burned north through the study area before turning west toward the town of South Park. Although exact records are not available, the wildfire burned several hundred thousand acres. In 1950, a wildfire started while land was being cleared for a golf course. As a result, nine fire fighters died and 92 buildings were destroyed. Since 1950 the area has not had any significant wildfires. In addition, the Pike National Forest has not conducted any prescribed fire or mechanical fuel treatments in the area. The Colorado Springs Fire Department recognized that individuals cannot be expected to make rational decisions about wildfire risk if they are not aware of the nature of the risk they face. The need to better inform its public about the risk of wildfire in the wildland–urban interface is clearly stated in the 2001 Colorado Springs Fire Department Wildfire Mitigation Plan: In general, the public does not perceive a risk from fire in the wildland– urban interface. Further, property owners believe that insurance companies or disaster assistance will always be there to cover losses.
Homebuyers and Wildfire Risk
61
When people believe the government will protect them from natural hazards, the damage potential of a catastrophic event increases. Fire prevention efforts, official pronouncements, and media depictions of imminent risk have been shown to have little effect on those in danger. The effects of public education efforts have not been significant when compared to the need. Unless a catastrophic event occurs, wildland=urban interface protection issues generate little interest. (6) In 2002, the Colorado Springs Fire Department began a unique project to rate the wildfire risk of 35,000 parcels in the wildland–urban interface and make the information available on a Web site. They believed that existing wildfire risk education efforts, which provided more general information, were ineffective, and that parcel-level wildfire risk assessments would provide the specific information needed to change homeowners’ behaviors. For each parcel, up to 25 variables were used to calculate an overall wildfire risk rating (low, medium, high, very high, or extreme). Although up to 25 variables are used, 6 variables largely determine a parcel’s wildfire risk rating. These are, in order of importance, proximity to dangerous topography, roof material, composition and abundance of combustible vegetation around the house, siding material, average slope of the parcel, and defensible space around the home. In June 2002 the fire department posted the parcel-level wildfire risk ratings on the Web (http:// csfd.springsgov.com). Homeowners could look up the wildfire risk rating of their house, or any other house, and receive information on how to mitigate wildfire risk. If homeowners take action to reduce the wildfire risk on their property, the fire department will reassess their wildfire risk rating. Since June 2002, the fire department has conducted several thousand reassessments. Although the Colorado Springs Fire Department’s FireWise program was not specifically targeting individuals moving into the area, anecdotal evidence suggested that some homebuyers were using the Web site to compare the wildfire risk ratings of prospective properties (personal communication with Cathy Prudhomme, Colorado Springs Fire Department, January 27, 2005). Since its launch, the average number of hits per day to the Colorado Springs wildfire risk rating Web site has increased every year from approximately 4,265 per day in 2002 to 12,858 per day in 2006. We cannot say that all these hits are coming from current and prospective homeowners in the Colorado Springs wildland– urban interface, but it is clear that the Web site has generated increasing interest.
A Hedonic Price Model Hedonic price models were estimated to examine the relationship between the sale price of a home and attributes of the home (Donovan, Champ, and Butry 2007; see Table 1 for variable descriptions). In addition to the usual attributes (square footage, type of construction, school district), the models included the wildfire risk ratings. The wildfire risk ratings collected by the Colorado Springs Fire Department were combined with house sales, housing characteristics, and neighborhood characteristics data obtained from the El Paso County Assessor’s office to estimate the hedonic price models. In the study area, 3,116 homes were sold between July 2002 (launch of the Web site) and September 2004. A typical house is 23 years old,
62
P. A. Champ et al.
Table 1. Definition of independent regression variables Variable CONDO DUPLEX FRAME TRACT MANSION AGE ROOMS BASEMENT ln(HOUSE) GARAGE H2 CS11 A20 ln(LOT) BUSY_MEDIUM BUSY_HIGH SALE_03 SALE_04 EXTREME VERY_HIGH HIGH MODERATE TOP_HIGH TOP_MEDIUM ROOF SIDING
Description Dummy variable for construction style (1 if condo, 0 otherwise) Dummy variable for construction style (1 if duplex, 0 otherwise) Dummy variable for construction type (1 if frame, 0 otherwise) Dummy variable for construction quality (1 if tract or low, 0 otherwise) Dummy variable for construction quality (1 if mansion, 0 otherwise) Year house was built subtracted from 2005 Number of rooms Finished basement square footage Natural log of total above ground square footage Garage square footage Dummy variable for school district (1 if Harrison 2, 0 otherwise) Dummy variable for school district (1 if Colorado Springs 11, 0 otherwise) Dummy variable for school district (1 if Academy 20, 0 otherwise) Natural log of lot square footage Dummy variable for traffic volume (1 if medium, 0 otherwise) Dummy variable for traffic volume (1 if high, 0 otherwise) Dummy variable for sale year (1 if 03, 0 otherwise) Dummy variable for sale year (1 if 04, 0 otherwise) Dummy variable for fire risk rating (1 if extreme, 0 otherwise) Dummy variable fire risk rating (1 if very high, 0 otherwise) Dummy variable for fire risk rating (1 if high, 0 otherwise) Dummy variable for fire risk rating (1 if moderate, 0 otherwise) Dummy variable for distance to dangerous topography (1 if < 30 feet, 0 otherwise) Dummy variable for distance to dangerous topography (1 if 30–100 feet, 0 otherwise) Dummy variable for roofing material (1 if wood, 0 otherwise) Dummy variable for siding material (1 if wood, 0 otherwise) (Continued )
Homebuyers and Wildfire Risk
63
Table 1. Continued Variable VEG_HIGH VEG_MED SLOPE
Description Dummy variable for veg. density within 30 feet of house (1 if dense, 0 otherwise) Dummy variable for veg. density within 30 feet of house (1 if moderately dense, 0 otherwise) Average slope (%) within 150 feet of house
Note. This table is Table 1 in Donovan, Champ, and Butry (2007) and is used with permission.
Table 2. Regression results for overall wildfire risk rating model (n ¼ 3116) Variable
Coefficient
Standard error
p Value
Marginal effect ($)a
CONSTANT CONDO DUPLEX FRAME TRACT MANSION AGE ROOMS BASEMENT ln(HOUSE) GARAGE H2 CS11 A20 ln(LOT) BUSY_MEDIUM BUSY_HIGH SALE_03 SALE_04 EXTREME VERY_HIGH HIGH MEDIUM R-squared q k
6.95 4.22E-02 6.80E-02 3.65E-02 0.160 2.04E-01 1.41E-03 4.53E-05 1.32E-04 0.433 1.33E-04 9.19E-02 5.85E-02 1.06E-01 4.76E-02 1.15E-02 1.97E-02 2.48E-02 8.37E-02 1.79E-02 2.18E-02 1.73E-02 6.70E-03 0.871 0.143 0.373
0.250 2.03E-02 2.70E-02 1.27E-02 1.24E-02 1.31E-02 2.45E-04 2.58E-03 5.70E-06 1.37E-02 1.89E-05 6.35E-02 1.44E-02 1.76E-01 3.70E-03 8.92E-03 1.06E-02 8.17E-03 9.00E-03 1.75E-02 1.52E-02 1.36E-02 1.26E-02
